AKSHAT PATIDAR
AI/ML Engineer · IIT Jodhpur
IIT Jodhpur
CGPA 8.54
live · your time
need an
ML engineer?
tear one ↓
ML engineer?
genuine ming · priceless
about me
hi, i'm akshat -
I build machine learning systems that have to survive contact with real people: RAG pipelines, forecasting models, computer-vision experiments and the data plumbing underneath them. Currently studying at IIT Jodhpur (8.54 CGPA) and working part-time on retrieval systems for a software company in Japan.
I care about the unglamorous parts - evaluation, failure modes, latency - because that's where "demo" becomes "product". My favourite compliment is "oh, it just works."
- akshat
@ IITJ
then simplify,
then ship
selected work
cooking… 🍳
the story so far
Qualitysoft · Japan
AI/ML Engineer
Building RAG pipelines with LangGraph and DeepEval-based QA gates. Wrote the framework we now use to compare orchestration stacks honestly.
flip ↻
field notes
"evals before vibes." the deepeval gate caught two silent regressions before any human noticed.
flip back ↻Government of Rajasthan
Data Science Intern
Data-backed planning for CT & MRI tenders across 30+ districts. CT turnaround time dropped 32% after the analysis shipped.
flip ↻
field notes
30+ districts of messy spreadsheets. the model was the easy part — the plumbing won it.
flip back ↻DevOPS FARM
AI Intern
Built an agentic AI psychologist with emotion recognition - 83% of 200 surveyed users preferred the bot-guided sessions.
flip ↻
field notes
83% preferred talking to the bot. still thinking about what that says about us.
flip back ↻IIT Jodhpur
Energy Analyst
LCOE models for renewable projects - realistic payback scenarios grounded in local context instead of brochure numbers.
flip ↻
field notes
brochure payback: 4 years. honest payback: 7. honesty won the argument.
flip back ↻tools i reach for
r · c
langchain · langgraph
llamaindex · deepeval
power bi · excel
rstudio
retrieval quality,
agent orchestration
say hello
building something ambitious?
akshatkhaptya@gmail.com ☏ +91 91099 12301if it involves machine learning, data systems or applied AI that has to actually ship - i want to hear about it. let's build something creative.
leave a note
on my board
visitor notes ↓